Abstract:

Traditional lace pattern retrieval mainly relies on the manual retrieval and text retrieval, while the text retrieval marks image by language description which leads to the unstable annotation information, low efficiency and unreliable retrieval results. This work marked image by texture pattern feature. A lace retrieval algorithm containing classification selection and classifier fusion through hierarchy match was proposed, and it made up the deficiency of the single matching method. Three image-based methods, such as gray level co-occurrence matrix, gray level-gradient co-occurrence matrix, local binary pattern operator, were fused by means of geometry features and invariant moments for a match-by-level respectively. Experimental results indicate that the performance of fusion-based method is better than any single method and it can achieve the lace retrieval promisingly as improving the reliability and accuracy of image retrieval effectively.